feat: Add support for ADAS project in labeling and export processes

- Implemented class name loading for ADAS campaigns in `_class_names_for_campaign`.
- Enhanced annotation parsing to support cuboid and 2D rectangle labels in `_parse_ls_annotations`.
- Updated catalog building to include ADAS packs and batches in `build_catalog_signature`.
- Added ADAS cuboid promotion logic in `adas_cuboid.py`, supporting both quaternion_json and YOLO HBB formats.
- Introduced validation for ADAS batches to check for YOLO HBB labels in `adas_cuboid.py`.
- Modified delivery scanning to include ADAS project deliveries in `scan.py`.
- Extended job execution to handle ADAS exports for YOLO format in `runner.py`.
- Updated labeling export logic to accommodate ADAS project in `export_cuboid_batch.py`.
- Enhanced format conversion to support rectangle labels in `format_converter.py`.
- Improved CVAT task handling in `service.py` to ensure task ID updates on each labeling job.
- Updated web API endpoints to reflect ADAS project changes in `hsap-api.ts`.
- Added ADAS catalog types and UI components for displaying ADAS packs and batches in `dmsCatalog.ts`, `CatalogPage.tsx`, and `ExportPage.tsx`.
- Adjusted workflow registry to align with new ADAS project structure.

.gitignore

@@ -49,6 +49,8 @@ datasets/**/inbox/
 datasets/**/archive/
 datasets/**/sources/
 datasets/**/DATASET/
+datasets/**/versions/
+datasets/**/manifests/ingest_log.jsonl
 datasets/**/*.jpg
 datasets/**/*.jpeg
 datasets/**/*.png

docs/ADAS_MOON3D_PACK.md

@@ -1,13 +1,18 @@
-# ADAS MOON-3D 训练包 `adas_moon3d_v1`
+# ADAS 前向训练包 `adas_v1`
+
+> **2026-07 更新**：pack 已从 `adas_moon3d_v1` 重命名为 `adas_v1`（见 `workflow.registry.yaml`）。
+> 新增 YOLO HBB 导出路径和 2D 矩形标注支持。
 
 ## 目录结构
 
 ```text
-datasets/adas/packs/adas_moon3d_v1/
+datasets/adas/packs/adas_v1/
 ├── sources/{batch}/
 │   ├── images/
 │   ├── calib/
-│   └── labels/quaternion_json/
+│   └── labels/
+│       ├── quaternion_json/   # 3D cuboid 导出（有 calib 时优先）
+│       └── yolo-hbb/          # 2D YOLO HBB 导出（无 calib 回退）
 ├── lists/
 │   ├── train_stems.txt
 │   └── val_stems.txt
@@ -32,20 +37,34 @@ datasets/adas/packs/adas_moon3d_v1/
 ## 管线
 
 1. **labeling_export** — CVAT ls_annotations → `labels/quaternion_json/*.json`
+   - 支持 **cuboid**（3D 框）和 **rectanglelabels**（2D 矩形）两种标注类型
+   - 2D 矩形自动转为 quaternion_json detection（`fit_ok=false`, `source="2d_rect"`）
 2. **cuboid_fit_3d**（有 calib 时自动触发）— 补全 3D 字段
 3. **build_adas**（审核）— `promote_batch` 复制到 pack + 刷新 stem 列表
+   - 优先使用 quaternion_json，无 cuboid 标注时回退到 YOLO HBB
+
+### 导出路径说明
+
+| 标注类型 | 导出产物 | 条件 |
+|---------|---------|------|
+| cuboid (3D) | `labels/quaternion_json/*.json` | CVAT cuboid 标注 |
+| rectanglelabels (2D) | `labels/quaternion_json/*.json` | 2D 矩形自动桥接 |
+| YOLO HBB | `labels/yolo-hbb/*.txt` | 无 cuboid 时使用 `export_adas_yolo` |
 
 ## CLI
 
 ```bash
-# 导出（平台 Job 或脚本）
+# cuboid 导出（平台 Job）
 PYTHONPATH=platform python3 -c "from as_platform.labeling.export_cuboid_batch import export_batch; ..."
 
+# YOLO HBB 导出（平台 Job，无 cuboid 时自动选择）
+# 由 jobs/runner.py 调用 datasets/adas/scripts/export_ls_to_yolo.py
+
 # 3D 拟合
 PYTHONPATH=platform python3 -c "from as_platform.labeling.fit_cuboid_batch import fit_batch; ..."
 
 # 入包
-python as.py build adas cuboid_7cls --batch val_front6mm_pilot --pack adas_moon3d_v1
+python as.py build adas cuboid_7cls --batch val_front6mm_pilot --pack adas_v1
 ```
 
 ## Smoke
@@ -54,7 +73,6 @@ python as.py build adas cuboid_7cls --batch val_front6mm_pilot --pack adas_moon3
 bash scripts/smoke_adas_promote.sh
 ```
 
-## 与 dms/packs/adas_v1 的区别
+## 数据目录
 
-- `dms/packs/adas_v1`：2D YOLO 历史包（[`scripts/organize_adas.py`](../scripts/organize_adas.py)）
+ADAS 前向数据在 **数据目录** → **前向** 视图中展示，按 pack → batch 层级显示图片数、标注数和类别分布。
-- `datasets/adas/packs/adas_moon3d_v1`：MOON-3D quaternion_json 3D GT

docs/DATA_LAKE_GAP.md

@@ -18,7 +18,7 @@
 
 - 数据候选：`platform/as_platform/db/models.py` → `DatasetCandidate`
 - 上传 API：`server.py` → `upload/file`、`inspect-upload`
-- 审核：`audit/queue.py`、`/api/v1/approvals/*`
+- 审核：`audit/queue.py`、`/api/v1/system/audit/*`
 - 入湖 CLI：`as.py build` / `add` + `ingest_incremental.py`
 
 ## 建议下一里程碑（未在本汇总 plan 全量实现）

docs/DMS_E2E_2IMG_TEST.md

@@ -119,7 +119,7 @@ bash scripts/smoke_dms_e2e_2img.sh setup
 ### Commit `0b8ade0` — Unified Ingest SDK
 
 - `platform/as_platform/data/promote/` — DMS/ADAS promote_batch
-- ADAS cuboid export / fit / `adas_moon3d_v1`
+- ADAS cuboid export / fit / `adas_v1`
 - `platform/as_platform/tests/test_unified_ingest_sdk.py`（7 项单元测试）
 - `scripts/smoke_adas_promote.sh`
 - `scripts/smoke_labeling_api.sh` 已接入上述测试

docs/HANDOVER.md

@@ -350,7 +350,7 @@ python as.py train dms dam --track platform  # 平台训练（需审核）
 | `/api/v1/system` | `system_routes.py` | 审核队列、任务监控、执行日志 |
 | `/api/v1/delivery` | `delivery_routes.py` | 批次交付、Bitable 集成 |
 | `/api/v1/feishu` | `feishu_routes.py` | Bitable Webhook 回调 |
-| `/api/v1/` (根) | `server.py` | health, dashboard, catalog, approvals, simulate, agents |
+| `/api/v1/` (根) | `server.py` | health, dashboard, catalog, simulate, agents |
 
 ### 6.2 核心端点速查
 
@@ -366,10 +366,10 @@ GET  /api/v1/catalog
 POST /api/v1/catalog
 
 # 审核流程
-GET  /api/v1/approvals
+GET  /api/v1/system/audit
-POST /api/v1/approvals/{id}/approve
+POST /api/v1/system/audit/{id}/approve
-POST /api/v1/approvals/{id}/reject
+POST /api/v1/system/audit/{id}/reject
-GET  /api/v1/approvals/{id}/preview
+GET  /api/v1/system/audit/{id}/preview
 
 # 任务队列
 GET  /api/v1/jobs

docs/LABELING_SOP.md

@@ -15,7 +15,7 @@
 
 **协调员 QA 门禁路径（推荐）：** 标完 → **提交质检** (`in_review`) → 质检通过 (`labeling_submitted`) → **执行导出** → `returned` → **提交 build**（审核）→ `ingested`。
 
-ADAS cuboid 详见 [ADAS_MOON3D_PACK.md](./ADAS_MOON3D_PACK.md)：导出 → 可选 `cuboid_fit_3d` → `build_adas` 进 `adas_moon3d_v1`。
+ADAS cuboid 详见 [ADAS_MOON3D_PACK.md](./ADAS_MOON3D_PACK.md)：导出（支持 cuboid 3D / 2D 矩形 / YOLO HBB 回退）→ 可选 `cuboid_fit_3d` → `build_adas` 进 `adas_v1`。
 
 **登记 meta**（`更多操作` 内）只写入 `batch.meta.yaml`，**不等于**已送标。
 
@@ -191,6 +191,7 @@ datasets/dms/inbox/dam/batch_0516/
 |------|--------|--------|------|
 | **训练包**（默认） | `dms_v1` 等 | `dam` / `addw_face` / `前向·粗检测` / `前向·细分类` … | 该包下各任务的 train/val 快照；前向在 registry 为单任务双 mode，目录按 mode 拆开显示 |
 | **采集批次** | 任务（如 `dam`） | `0516 批次` 等 mode | 单波送标质量 |
+| **前向** | `adas_v1` 等 pack | 批次列表 | ADAS 3D/2D 标注批次，含图片数、标注数、类别分布；支持 quaternion_json 和 YOLO HBB |
 | **车道线** | lane pack | — | Lane 列表统计 |
 
 增量条（合并前）仍在磁盘 `packs/<pack>/<task>/sources/<批次名>/`；合并用 `python HSAP/as.py build dms dam --pack dms_v1 --all-sources`。

platform/as_platform/audit/review.py

@@ -68,7 +68,13 @@ def _class_names_for_campaign(camp) -> dict[int, str]:
     import yaml
     from as_platform.data.core import load_wf, proj_root
 
-    if not camp or camp.project != "dms":
+    if not camp:
+        return {}
+    if camp.project == "adas":
+        from as_platform.labeling.class_map import load_adas_class_names
+        names = load_adas_class_names()
+        return {i: n for i, n in enumerate(names)}
+    if camp.project != "dms":
         return {}
     wf = load_wf()
     root = proj_root(wf, "dms")
@@ -114,21 +120,42 @@ def _parse_ls_annotations(path: Path, class_names: dict[int, str]) -> list[dict[
         return []
     out: list[dict[str, Any]] = []
     for item in data.get("result") or []:
-        if item.get("type") not in ("rectanglelabels", "rectangle"):
+        item_type = item.get("type")
-            continue
+        if item_type == "cuboid":
-        val = item.get("value") or {}
+            # Cuboid: 8 vertices in pixel coords → compute 2D bbox
-        w_pct = float(val.get("width") or 0)
+            points = item.get("points") or []
-        h_pct = float(val.get("height") or 0)
+            if len(points) < 16:
-        if w_pct <= 0 or h_pct <= 0:
+                continue
-            continue
+            label = item.get("label") or "unknown"
-        x_pct = float(val.get("x") or 0)
+            cid = _name_to_class_id(str(label), class_names)
-        y_pct = float(val.get("y") or 0)
+            # Compute min/max from 8 (x,y) pairs
-        labels = val.get("rectanglelabels") or val.get("labels") or []
+            xs = [points[i] for i in range(0, len(points), 2)]
-        label = labels[0] if labels else "unknown"
+            ys = [points[i] for i in range(1, len(points), 2)]
-        cid = _name_to_class_id(str(label), class_names)
+            min_x, max_x = min(xs), max(xs)
-        cx = (x_pct + w_pct / 2) / 100.0
+            min_y, max_y = min(ys), max(ys)
-        cy = (y_pct + h_pct / 2) / 100.0
+            if max_x <= min_x or max_y <= min_y:
-        out.append({"class_id": cid, "bbox": (cx, cy, w_pct / 100.0, h_pct / 100.0)})
+                continue
+            orig_w = float(item.get("original_width") or 1920)
+            orig_h = float(item.get("original_height") or 1080)
+            bw = (max_x - min_x) / orig_w
+            bh = (max_y - min_y) / orig_h
+            cx = (min_x + max_x) / 2.0 / orig_w
+            cy = (min_y + max_y) / 2.0 / orig_h
+            out.append({"class_id": cid, "bbox": (cx, cy, bw, bh)})
+        elif item_type in ("rectanglelabels", "rectangle"):
+            val = item.get("value") or {}
+            w_pct = float(val.get("width") or 0)
+            h_pct = float(val.get("height") or 0)
+            if w_pct <= 0 or h_pct <= 0:
+                continue
+            x_pct = float(val.get("x") or 0)
+            y_pct = float(val.get("y") or 0)
+            labels = val.get("rectanglelabels") or val.get("labels") or []
+            label = labels[0] if labels else "unknown"
+            cid = _name_to_class_id(str(label), class_names)
+            cx = (x_pct + w_pct / 2) / 100.0
+            cy = (y_pct + h_pct / 2) / 100.0
+            out.append({"class_id": cid, "bbox": (cx, cy, w_pct / 100.0, h_pct / 100.0)})
     return out
 
 

platform/as_platform/data/catalog_cache.py

@@ -83,7 +83,7 @@ def build_catalog_signature(wf: dict, proj_root_fn) -> dict[str, Any]:
             files.append({"path": str(p), "missing": True})
 
     dirs: list[dict[str, Any]] = []
-    for pname in ("dms", "lane"):
+    for pname in ("dms", "lane", "adas"):
         root = proj_root_fn(wf, pname)
         dirs.append(_dir_fingerprint(root, scan_children=False))
         dirs.append(_dir_fingerprint(root / "inbox"))

platform/as_platform/data/core.py

@@ -11,7 +11,7 @@ from typing import Any
 import yaml
 
 from as_platform.config import WORKSPACE, WORKSPACE_ROOT, LANE_DATA_VIZ_ENABLED
-from as_platform.data.batch import META_FILENAME, dms_has_images, enrich_batch, write_meta
+from as_platform.data.batch import META_FILENAME, dms_has_images, enrich_batch, read_meta, write_meta
 from as_platform.data.catalog_cache import (
     build_catalog_signature,
     get_cached_catalog,
@@ -811,7 +811,7 @@ def _normalize_catalog_dms(dms: dict[str, Any], reg: dict) -> dict[str, Any]:
 
 
 def _build_catalog(wf: dict, *, prefer_reports: bool = True) -> tuple[dict[str, Any], str]:
-    out: dict[str, Any] = {"workspace": str(WORKSPACE), "dms": {}, "lane": {}}
+    out: dict[str, Any] = {"workspace": str(WORKSPACE), "dms": {}, "lane": {}, "adas": {}}
     build_source = "scan"
 
     reports = load_dms_reports()
@@ -966,6 +966,71 @@ def _build_catalog(wf: dict, *, prefer_reports: bool = True) -> tuple[dict[str,
             "add_template": "python as.py add lane --src <archive> --engineer <name> --date YYYYMMDD",
             "quality": lane_quality,
         }
+
+    # ── ADAS catalog ──
+    adas_root = proj_root(wf, "adas")
+    out["adas"] = {}
+    adas_packs_dir = adas_root / "packs"
+    if adas_packs_dir.is_dir():
+        for pack_dir in sorted(adas_packs_dir.iterdir()):
+            if pack_dir.name.startswith(".") or not pack_dir.is_dir():
+                continue
+            pack_name = pack_dir.name
+            sources_dir = pack_dir / "sources"
+            pack_entry: dict[str, Any] = {
+                "name": pack_name,
+                "path": str(pack_dir.relative_to(adas_root)),
+                "enabled": pack_name in wf["projects"]["adas"].get("active_packs", []),
+                "batches": [],
+                "total_images": 0,
+                "total_labels": 0,
+                "class_counts": {},
+            }
+            if sources_dir.is_dir():
+                for batch_dir in sorted(sources_dir.iterdir()):
+                    if batch_dir.name.startswith(".") or not batch_dir.is_dir():
+                        continue
+                    batch_name = batch_dir.name
+                    meta = read_meta(batch_dir) or {}
+                    stage = meta.get("stage", "unknown")
+                    task = meta.get("task", "")
+                    counts = meta.get("counts", {})
+                    img_count = counts.get("images", 0)
+                    # Count quaternion_json detections
+                    qdir = batch_dir / "labels" / "quaternion_json"
+                    label_count = 0
+                    class_counts: dict[str, int] = {}
+                    if qdir.is_dir():
+                        for qf in sorted(qdir.glob("*.json")):
+                            try:
+                                qdata = json.loads(qf.read_text(encoding="utf-8"))
+                            except (OSError, json.JSONDecodeError):
+                                continue
+                            for det in qdata.get("detections") or []:
+                                label_count += 1
+                                cls = str(det.get("class_name") or det.get("class_id", "unknown"))
+                                class_counts[cls] = class_counts.get(cls, 0) + 1
+                    # Count images from directory if meta is missing
+                    if img_count == 0:
+                        imgs_dir = batch_dir / "images"
+                        if imgs_dir.is_dir():
+                            img_count = sum(1 for f in imgs_dir.rglob("*") if f.suffix.lower() in IMG_EXTS)
+                    batch_entry = {
+                        "batch": batch_name,
+                        "task": task,
+                        "stage": stage,
+                        "images": img_count,
+                        "labels": label_count,
+                        "class_counts": class_counts,
+                        "ingested_at": meta.get("ingested_at"),
+                    }
+                    pack_entry["batches"].append(batch_entry)
+                    pack_entry["total_images"] += img_count
+                    pack_entry["total_labels"] += label_count
+                    for cls, n in class_counts.items():
+                        pack_entry["class_counts"][cls] = pack_entry["class_counts"].get(cls, 0) + n
+            out["adas"][pack_name] = pack_entry
+
     return out, build_source
 
 
@@ -996,7 +1061,9 @@ def get_catalog(
         result = {"task": task_or_pack, **(full_catalog.get("dms", {}).get(task_or_pack, {}))}
     elif project == "lane" and task_or_pack:
         result = {"pack": task_or_pack, **(full_catalog.get("lane", {}).get(task_or_pack, {}))}
-    elif project in ("dms", "lane"):
+    elif project == "adas" and task_or_pack:
+        result = {"pack": task_or_pack, **(full_catalog.get("adas", {}).get(task_or_pack, {}))}
+    elif project in ("dms", "lane", "adas"):
         result = {"workspace": full_catalog.get("workspace", str(WORKSPACE)), project: full_catalog.get(project, {})}
     else:
         result = dict(full_catalog)

platform/as_platform/data/promote/adas_cuboid.py

@@ -86,14 +86,20 @@ class AdasCuboidPromoteAdapter(PackPromoteAdapter):
     def promote(self, ctx: PromoteContext) -> PromoteResult:
         warnings = list(ctx.extra.get("validate_warnings") or [])
         qdir = ctx.batch_dir / "labels" / "quaternion_json"
-        if not qdir.is_dir() or not any(qdir.glob("*.json")):
+        has_quat = qdir.is_dir() and any(qdir.glob("*.json"))
+
+        # YOLO HBB 回退
+        labels_dir = ctx.batch_dir / "labels"
+        has_yolo = (labels_dir / "yolo-hbb").is_dir() and any((labels_dir / "yolo-hbb").glob("*.txt"))
+
+        if not has_quat and not has_yolo:
             return PromoteResult(
                 ok=False,
                 project=ctx.project,
                 task=ctx.task,
                 batch=ctx.batch,
                 pack=ctx.pack,
-                warnings=["missing quaternion_json export"],
+                warnings=["missing quaternion_json or YOLO HBB export"],
             )
 
         pack_dir = ctx.project_root / "packs" / ctx.pack
@@ -119,7 +125,9 @@ class AdasCuboidPromoteAdapter(PackPromoteAdapter):
             if src_sub.is_dir():
                 copied += _sync_tree(src_sub, dest / sub)
 
-        normalized = _normalize_quaternion_json(dest)
+        normalized = _normalize_quaternion_json(dest) if has_quat else 0
+        if has_yolo and not has_quat:
+            warnings.append("using YOLO HBB labels (no quaternion_json)")
 
         meta = read_meta(ctx.batch_dir) or {}
         meta.update({
@@ -130,6 +138,7 @@ class AdasCuboidPromoteAdapter(PackPromoteAdapter):
             "pack": ctx.pack,
             "ingested_at": datetime.now(timezone.utc).isoformat(),
             "pipeline_version": 2,
+            "label_format": "quaternion_json" if has_quat else "yolo_hbb",
         })
         write_meta(dest, meta)
         write_meta(ctx.batch_dir, meta)

platform/as_platform/data/promote/validate/adas_cuboid.py

@@ -1,4 +1,4 @@
-"""ADAS cuboid batch validation before promote."""
+"""ADAS batch validation before promote（支持 quaternion_json 和 YOLO HBB）。"""
 from __future__ import annotations
 
 import json
@@ -7,20 +7,87 @@ from pathlib import Path
 from as_platform.labeling.class_map import load_adas_class_names
 
 
+def _has_yolo_hbb_labels(batch_dir: Path) -> bool:
+    """检查是否有 YOLO HBB txt 标注文件。"""
+    yolo_dir = batch_dir / "labels" / "yolo-hbb"
+    return yolo_dir.is_dir() and any(yolo_dir.glob("*.txt"))
+
+
+def _validate_yolo_hbb(batch_dir: Path, expected_names: list[str]) -> tuple[list[str], list[str], dict]:
+    """校验 YOLO HBB 格式标注。"""
+    errors: list[str] = []
+    warnings: list[str] = []
+    n_classes = len(expected_names)
+
+    labels_dir = batch_dir / "labels" / "yolo-hbb"
+    txt_files = sorted(labels_dir.glob("*.txt")) if labels_dir.is_dir() else []
+
+    total_boxes = 0
+    files_with_boxes = 0
+    for p in txt_files:
+        try:
+            lines = p.read_text(encoding="utf-8").strip().splitlines()
+        except OSError as e:
+            errors.append(f"{p.name}: read error ({e})")
+            continue
+        if not lines:
+            continue
+        has_valid = False
+        for line in lines:
+            line = line.strip()
+            if not line:
+                continue
+            parts = line.split()
+            if len(parts) < 5:
+                warnings.append(f"{p.name}: invalid YOLO line '{line[:40]}...'")
+                continue
+            try:
+                cid = int(parts[0])
+                if cid < 0 or cid >= n_classes:
+                    errors.append(f"{p.name}: class_id {cid} out of range (0-{n_classes - 1})")
+                    continue
+                # 验证 bbox 值在 [0,1]
+                for v in parts[1:5]:
+                    fv = float(v)
+                    if fv < 0 or fv > 1:
+                        warnings.append(f"{p.name}: bbox value {fv} out of [0,1]")
+            except ValueError:
+                warnings.append(f"{p.name}: non-numeric value in '{line[:40]}...'")
+                continue
+            total_boxes += 1
+            has_valid = True
+        if has_valid:
+            files_with_boxes += 1
+
+    if files_with_boxes == 0:
+        errors.append("no valid YOLO HBB label files")
+
+    stats = {
+        "format": "yolo_hbb",
+        "label_files": len(txt_files),
+        "files_with_boxes": files_with_boxes,
+        "total_boxes": total_boxes,
+    }
+    return errors, warnings, stats
+
+
 def validate_adas_cuboid_batch(
     batch_dir: Path,
     *,
     allow_partial_3d: bool = False,
     min_fit_ratio: float = 0.8,
 ) -> tuple[list[str], list[str], dict]:
-    """Return (errors, warnings, stats)."""
+    """Return (errors, warnings, stats). 优先 quaternion_json，其次 YOLO HBB。"""
     errors: list[str] = []
     warnings: list[str] = []
     qdir = batch_dir / "labels" / "quaternion_json"
     expected_names = load_adas_class_names()
 
     if not qdir.is_dir():
-        errors.append(f"missing labels/quaternion_json under {batch_dir}")
+        # 回退到 YOLO HBB 格式
+        if _has_yolo_hbb_labels(batch_dir):
+            return _validate_yolo_hbb(batch_dir, expected_names)
+        errors.append(f"missing labels/quaternion_json or YOLO HBB labels under {batch_dir}")
         return errors, warnings, {}
 
     files = sorted(qdir.glob("*.json"))

platform/as_platform/deliveries/scan.py

@@ -34,7 +34,13 @@ def _scan_project_inbox(project: str, wf: dict | None = None) -> list[dict[str,
 
     with session_scope() as db:
         deliveries = {
-            (r.project, r.task or "", r.mode or "", r.batch_name): r
+            (r.project, r.task or "", r.mode or "", r.batch_name): {
+                "id": r.id,
+                "status": r.status,
+                "collection_start": r.collection_start,
+                "collection_end": r.collection_end,
+                "created_at": r.created_at,
+            }
             for r in db.query(BatchDelivery).filter(BatchDelivery.project == project).all()
         }
         indexed = {
@@ -78,13 +84,13 @@ def _scan_project_inbox(project: str, wf: dict | None = None) -> list[dict[str,
                 "has_labels": has_labels,
                 "stage_hint": stage_hint,
                 "source_type": "inbox_scan",
-                "delivery_id": delivery.id if delivery else None,
+                "delivery_id": delivery["id"] if delivery else None,
-                "delivery_status": delivery.status if delivery else None,
+                "delivery_status": delivery["status"] if delivery else None,
                 "in_ledger": delivery is not None,
                 "in_workbench": in_index,
-                "collection_start": delivery.collection_start if delivery else _dir_mtime_iso(batch_dir),
+                "collection_start": delivery["collection_start"] if delivery else _dir_mtime_iso(batch_dir),
-                "collection_end": delivery.collection_end if delivery else None,
+                "collection_end": delivery["collection_end"] if delivery else None,
-                "created_at": delivery.created_at.isoformat() if delivery and delivery.created_at else None,
+                "created_at": delivery["created_at"].isoformat() if delivery and delivery["created_at"] else None,
                 "needs_ledger": delivery is None,
                 "needs_workbench": not in_index,
             })

platform/as_platform/jobs/runner.py

@@ -277,6 +277,24 @@ def execute_action(action: str, params: dict[str, Any]) -> dict[str, Any]:
                     f"skipped_empty={conv.get('skipped_empty')} missing_ann={conv.get('missing_ann')}"
                 )
             return {"ok": True, "stdout": json.dumps(conv, ensure_ascii=False), "stderr": "", "export_convert": conv}
+        if row.get("project") == "adas" and export == "yolo":
+            scripts_dir = WORKSPACE / "datasets" / "adas" / "scripts"
+            if str(scripts_dir) not in sys.path:
+                sys.path.insert(0, str(scripts_dir))
+            from export_ls_to_yolo import export_batch as export_adas_yolo
+
+            with session_scope() as db:
+                camp = db.get(LabelingCampaign, campaign_id)
+                if not camp:
+                    raise ValueError("campaign not found")
+                batch_dir = resolve_campaign_batch_dir(camp)
+            conv = export_adas_yolo(batch_dir, task=task, out_subdir="labels/yolo-hbb")
+            if conv.get("written", 0) == 0:
+                raise ValueError(
+                    "export_adas_yolo: 无有效 YOLO HBB 标注可导出 (written=0); "
+                    f"skipped_empty={conv.get('skipped_empty')} missing_ann={conv.get('missing_ann')}"
+                )
+            return {"ok": True, "stdout": json.dumps(conv, ensure_ascii=False), "stderr": "", "export_convert": conv}
         if row.get("project") == "adas" and export == "cvat_cuboid":
             from as_platform.labeling.export_cuboid_batch import export_batch as export_cuboid_batch
 

platform/as_platform/labeling/batch_stage.py

@@ -119,8 +119,8 @@ def on_labeling_export_job_succeeded(job: dict) -> None:
         except Exception:
             return
         project = camp.project or "dms"
-    if _batch_has_export_labels(project, batch_dir):
+    # job 已成功，无论是否有 CLI 导出产物，均进入 returned 阶段
-        _advance_campaign_stage(str(cid), "returned")
+    _advance_campaign_stage(str(cid), "returned")
     if project == "adas" and _batch_has_calib(batch_dir):
         from as_platform.jobs.queue import enqueue_job
 

platform/as_platform/labeling/export_cuboid_batch.py

@@ -102,8 +102,45 @@ def _resolve_image_for_ann(data: dict[str, Any], batch_dir: Path, task_id: str)
     return None
 
 
+def _rect_item_to_detection(item: dict[str, Any], class_map: dict[str, int]) -> dict[str, Any] | None:
+    """2D rectanglelabels 条目 → quaternion_json detection（无 3D 信息，仅 bbox2d）。"""
+    value = item.get("value") or {}
+    w_pct = float(value.get("width") or 0)
+    h_pct = float(value.get("height") or 0)
+    if w_pct <= 0 or h_pct <= 0:
+        return None
+    labels = value.get("rectanglelabels") or []
+    label = str(labels[0]) if labels else "unknown"
+    class_id = class_map.get(label)
+    if class_id is None:
+        for name, cid in class_map.items():
+            if name.lower() == label.lower():
+                class_id = cid
+                break
+    if class_id is None:
+        return None
+
+    img_w = int(item.get("original_width") or 1920)
+    img_h = int(item.get("original_height") or 1080)
+    x_pct = float(value.get("x") or 0)
+    y_pct = float(value.get("y") or 0)
+    x1 = int(x_pct / 100.0 * img_w)
+    y1 = int(y_pct / 100.0 * img_h)
+    x2 = int((x_pct + w_pct) / 100.0 * img_w)
+    y2 = int((y_pct + h_pct) / 100.0 * img_h)
+
+    return {
+        "class_id": class_id,
+        "class_name": label,
+        "score": 1.0,
+        "box2d_xyxy": [x1, y1, x2, y2],
+        "fit_ok": False,
+        "source": "2d_rect",
+    }
+
+
 def export_batch(batch_dir: Path) -> dict[str, Any]:
-    """导出 cuboid ls_annotations → quaternion_json。"""
+    """导出 ls_annotations（cuboid + 2D rectangle）→ quaternion_json。"""
     batch_dir = batch_dir.resolve()
     class_map = _load_cuboid_class_map()
     calib_path, K, calib_size = _find_calib(batch_dir)
@@ -124,7 +161,9 @@ def export_batch(batch_dir: Path) -> dict[str, Any]:
             continue
         regions = _extract_result_regions(data)
         cuboids = [r for r in regions if r.get("type") == "cuboid"]
-        if not cuboids:
+        rectangles = [r for r in regions if r.get("type") == "rectanglelabels"]
+
+        if not cuboids and not rectangles:
             skipped_empty += 1
             continue
 
@@ -134,16 +173,24 @@ def export_batch(batch_dir: Path) -> dict[str, Any]:
             continue
 
         detections: list[dict[str, Any]] = []
+        # 处理 cuboid 标注
         for item in cuboids:
             det = cuboid_item_to_detection(item, class_map, K=K)
             if det:
                 detections.append(det)
+        # 处理 2D 矩形标注 → quaternion_json detection
+        for item in rectangles:
+            det = _rect_item_to_detection(item, class_map)
+            if det:
+                detections.append(det)
         if not detections:
             skipped_empty += 1
             continue
 
-        img_w = int((cuboids[0].get("original_width") or (calib_size or [1920, 1080])[0]))
+        # 获取图片尺寸：优先 cuboid，其次 rectangle
-        img_h = int((cuboids[0].get("original_height") or (calib_size or [1920, 1080])[1]))
+        first_region = cuboids[0] if cuboids else rectangles[0]
+        img_w = int((first_region.get("original_width") or (calib_size or [1920, 1080])[0]))
+        img_h = int((first_region.get("original_height") or (calib_size or [1920, 1080])[1]))
 
         payload: dict[str, Any] = {
             "image": str(image_path),

platform/as_platform/labeling/format_converter.py

@@ -238,10 +238,14 @@ def cvat_shapes_to_export_regions(
         }
 
         if stype == "rectangle":
-            xtl = float(shape.get("xtl", 0))
+            pts = shape.get("points") or []
-            ytl = float(shape.get("ytl", 0))
+            if len(pts) >= 4:
-            xbr = float(shape.get("xbr", 0))
+                xtl, ytl, xbr, ybr = float(pts[0]), float(pts[1]), float(pts[2]), float(pts[3])
-            ybr = float(shape.get("ybr", 0))
+            else:
+                xtl = float(shape.get("xtl", 0))
+                ytl = float(shape.get("ytl", 0))
+                xbr = float(shape.get("xbr", 0))
+                ybr = float(shape.get("ybr", 0))
             regions.append({
                 **base,
                 "type": "rectanglelabels",

platform/as_platform/labeling/service.py

@@ -186,9 +186,9 @@ def open_campaign(
         else:
             camp.status = "in_progress"
             camp.updated_at = now
-            if cvat_task_id and not camp.cvat_task_id:
+            # 每次开标都创建新 CVAT Task，始终更新 task_id
-                camp.cvat_task_id = cvat_task_id
+            camp.cvat_task_id = cvat_task_id
-                camp.cvat_job_url = cvat_job_url
+            camp.cvat_job_url = cvat_job_url
             if ann_types and not camp.annotation_types:
                 camp.annotation_types = ann_types
         db.flush()
@@ -531,6 +531,7 @@ def _cvat_upload_thread(cvat_task_id: int, image_paths: list, campaign_id: str |
                     with session_scope() as db:
                         camp = db.get(LabelingCampaign, campaign_id)
                         if camp and camp.cvat_job_url is None:
+                            camp.cvat_task_id = cvat_task_id
                             camp.cvat_job_url = f"_UPLOAD_FAILED_: {e}"
                             camp.status = "upload_failed"
                             updated = True
@@ -542,12 +543,13 @@ def _cvat_upload_thread(cvat_task_id: int, image_paths: list, campaign_id: str |
                 try:
                     with session_scope() as db:
                         camp = db.get(LabelingCampaign, campaign_id)
-                        # 上传成功后，尝试回填 job_url
+                        # 上传成功后，回填 job_url 和 task_id
                         if camp and camp.cvat_job_url is None:
                             from as_platform.labeling.cvat_client import get_cvat_client
                             cvat = get_cvat_client()
                             task = cvat.get_task(cvat_task_id)
                             if task.job_url:
+                                camp.cvat_task_id = cvat_task_id
                                 camp.cvat_job_url = task.job_url
                                 camp.status = "in_progress"
                                 updated = True
@@ -657,7 +659,11 @@ def _build_class_map(camp, reg: dict | None) -> dict[str, int]:
 
 
 def get_cvat_status(campaign_id: str) -> dict[str, Any]:
-    """查询 CVAT 侧 Task 状态，包含上传进度信息。"""
+    """查询 CVAT 侧 Task 状态，包含上传进度信息。
+    
+    自动恢复：当检测到 cvat_job_url 为 None 且 CVAT Task 无数据时，
+    说明上传线程已丢失（如容器重启），自动重新触发上传。
+    """
     with session_scope() as db:
         camp = db.get(LabelingCampaign, campaign_id)
         if not camp:
@@ -690,12 +696,31 @@ def get_cvat_status(campaign_id: str) -> dict[str, Any]:
                 images = _iter_batch_images(batch_dir)
                 image_count = len(images)
             except Exception:
+                images = []
                 pass
+
+            # 自动恢复：cvat_job_url 为空且 CVAT Task 无 Job → 上传线程已丢失
+            resolved_url = task.job_url or camp.cvat_job_url
+            if not resolved_url and image_count > 0:
+                # CVAT Task 存在但无数据，重新触发上传
+                import threading
+                _needs_recovery = True
+                try:
+                    r = cvat._session.get(f"{cvat._api}/tasks/{camp.cvat_task_id}/data/meta")
+                    _needs_recovery = r.status_code >= 400  # 400 = 未上传数据
+                except Exception:
+                    pass
+                if _needs_recovery:
+                    print(f"[CVAT] 检测到断线任务，自动恢复上传 task={camp.cvat_task_id} campaign={campaign_id}")
+                    uploader = _cvat_upload_thread(camp.cvat_task_id, images, campaign_id=campaign_id)
+                    threading.Thread(target=uploader, daemon=True).start()
+                    camp.cvat_job_url = None  # 重置，等待线程回填
+
             return {
                 "cvat_available": True,
                 "campaign_id": campaign_id,
                 "cvat_task_id": camp.cvat_task_id,
-                "cvat_job_url": task.job_url or camp.cvat_job_url,
+                "cvat_job_url": resolved_url,
                 "cvat_status": task.status,
                 "image_count": image_count,
             }

platform/web/src/app/hsap-api.ts

@@ -396,18 +396,18 @@ export const hsapApi = {
     if (opts?.status) p.set("status", opts.status);
     if (opts?.offset != null) p.set("offset", String(opts.offset));
     if (opts?.limit != null) p.set("limit", String(opts.limit));
-    return fetchJson<PagedResult<Record<string, unknown>>>(`${API_BASE}/api/v1/approvals?${p}`);
+    return fetchJson<PagedResult<Record<string, unknown>>>(`${API_BASE}/api/v1/system/audit?${p}`);
   },
 
-  getApproval: (id: string) => fetchJson<Record<string, unknown>>(`${API_BASE}/api/v1/approvals/${id}`),
+  getApproval: (id: string) => fetchJson<Record<string, unknown>>(`${API_BASE}/api/v1/system/audit/${id}`),
-  getApprovalPreview: (id: string) => fetchJson<Record<string, unknown>>(`${API_BASE}/api/v1/approvals/${id}/preview`),
+  getApprovalPreview: (id: string) => fetchJson<Record<string, unknown>>(`${API_BASE}/api/v1/system/audit/${id}/preview`),
 
   listApprovalImages: (id: string, offset = 0, limit = 60) =>
-    fetchJson<{ total: number; items: { id: string }[] }>(`${API_BASE}/api/v1/approvals/${id}/images?offset=${offset}&limit=${limit}`),
+    fetchJson<{ total: number; items: { id: string }[] }>(`${API_BASE}/api/v1/system/audit/${id}/images?offset=${offset}&limit=${limit}`),
 
   fetchApprovalImageBlob: async (approvalId: string, imageId: string, thumb = true) => {
     const q = thumb ? "?thumb=true" : "?thumb=false";
-    const res = await fetch(`${API_BASE}/api/v1/approvals/${approvalId}/images/${imageId}${q}`, { headers: authHeaders(), cache: "no-store" });
+    const res = await fetch(`${API_BASE}/api/v1/system/audit/${approvalId}/images/${imageId}${q}`, { headers: authHeaders(), cache: "no-store" });
     if (!res.ok) throw new Error(await res.text());
     return URL.createObjectURL(await res.blob());
   },

platform/web/src/lib/dmsCatalog.ts

@@ -2,6 +2,7 @@
 export type CatalogReport = Record<string, unknown> & {
   dms?: Record<string, DmsTaskEntry>;
   lane?: Record<string, Record<string, unknown>>;
+  adas?: Record<string, AdasPackEntry>;
   projects?: { dms?: { active_packs?: string[] }; lane?: { active_packs?: string[] } };
   _cache?: Record<string, unknown>;
 };
@@ -19,6 +20,28 @@ export type DmsPackRow = {
   total_boxes?: number; label_files?: number; sampled?: boolean;
 };
 
+// ── ADAS catalog types ──
+
+export type AdasPackEntry = {
+  name: string;
+  path: string;
+  enabled: boolean;
+  total_images: number;
+  total_labels: number;
+  class_counts: Record<string, number>;
+  batches: AdasBatchEntry[];
+};
+
+export type AdasBatchEntry = {
+  batch: string;
+  task: string;
+  stage: string;
+  images: number;
+  labels: number;
+  class_counts: Record<string, number>;
+  ingested_at?: string;
+};
+
 export type SplitCounts = { train: number; val: number; test: number };
 
 export function aggregateSplitCounts(packs: DmsPackRow[]): SplitCounts {

platform/web/src/modules/labeling/components/ExportBatchTable.tsx

@@ -10,6 +10,7 @@ import type { LabelingBatchRow } from "@/lib/types";
 type ExportBatchTableProps = {
   batches: LabelingBatchRow[];
   importingId: string | null;
+  exportingId: string | null;
   buildingId: string | null;
   onExport: (campaignId: string) => void;
   onImportVendor: (campaignId: string) => void;
@@ -21,6 +22,7 @@ const COLS = ["34%", "14%", "8%", "10%", "auto", "7.5rem"];
 export const ExportBatchTable: React.FC<ExportBatchTableProps> = ({
   batches,
   importingId,
+  exportingId,
   buildingId,
   onExport,
   onImportVendor,
@@ -67,7 +69,7 @@ export const ExportBatchTable: React.FC<ExportBatchTableProps> = ({
             <td className="py-2 px-2 whitespace-nowrap w-[7.5rem]">
               {cid && isExport && (
                 <div className="inline-flex items-center justify-center gap-0.5">
-                  <Button size="small" variant="primary" className="!px-2" onClick={() => onExport(cid)}>
+                  <Button size="small" variant="primary" className="!px-2" loading={exportingId === cid} onClick={() => onExport(cid)}>
                     导出
                   </Button>
                   <Button

platform/web/src/modules/labeling/pages/CatalogPage.tsx

@@ -13,7 +13,8 @@ import {
   buildDmsPackTree, dmsPacks, dmsTaskModes,
   findPackRow, isDmsCatalogScope, packTaskKey, parseCatalogScope,
   parsePackTaskKey, primaryPack, scopeKeyFromSelection, selectionFromScopeKey,
-  splitCountsFromPack, type CatalogReport, type CatalogUiSelection,
+  splitCountsFromPack, type AdasBatchEntry, type AdasPackEntry,
+  type CatalogReport, type CatalogUiSelection,
   type CatalogViewKind, type DmsPackRow, type DmsTaskEntry,
 } from "@/lib/dmsCatalog";
 
@@ -183,8 +184,8 @@ export const CatalogPage: React.FC = () => {
             ))}
           </div>
 
-          {/* DMS / Forward sub-views */}
+          {/* DMS sub-views */}
-          {domain !== "lane" && (
+          {domain === "dms" && (
             <div className="flex flex-wrap gap-2 mb-4 text-sm">
               <p className="text-xs text-gray-500 m-0 self-center mr-2">视图</p>
               {(["pack", "batch"] as const).map((v) => (
@@ -198,8 +199,8 @@ export const CatalogPage: React.FC = () => {
             </div>
           )}
 
-          {/* Pack selector */}
+          {/* Pack selector — only for DMS */}
-          {domain !== "lane" && subView === "pack" && packTree.length > 0 && (
+          {domain === "dms" && subView === "pack" && packTree.length > 0 && (
             <div className="flex flex-wrap gap-3 items-end mb-4">
               <label className="flex-1 min-w-[8rem] text-xs text-gray-500">训练包
                 <select className={selectClass} value={ui.pack} onChange={(e) => setPack(e.target.value)}>
@@ -214,8 +215,75 @@ export const CatalogPage: React.FC = () => {
             </div>
           )}
 
-          {/* DMS/Forward: Charts */}
+          {/* ADAS 前向：Pack + Batch 表格 */}
-          {domain !== "lane" && taskEntry && isDmsCatalogScope(scope) && (
+          {domain === "forward" && cat?.adas && Object.keys(cat.adas).length > 0 && (
+            <div className="space-y-4">
+              {Object.entries(cat.adas as Record<string, AdasPackEntry>).map(([packName, pack]) => (
+                <div key={packName} className="rounded-xl border border-gray-200 bg-white overflow-hidden">
+                  <div className="bg-gray-50 px-4 py-3 border-b border-gray-200 flex items-center gap-3">
+                    <span className="font-semibold text-base">{packName}</span>
+                    {pack.enabled && <Badge variant="success">启用</Badge>}
+                    <span className="text-sm text-gray-500 ml-auto">
+                      {pack.total_images} 图 · {pack.total_labels} 标注 · {pack.batches.length} 批次
+                    </span>
+                  </div>
+                  {pack.batches.length > 0 ? (
+                    <table className="w-full text-sm">
+                      <thead>
+                        <tr className="bg-gray-50 text-left border-b border-gray-200">
+                          <th className="px-4 py-2">批次</th>
+                          <th className="px-4 py-2">任务</th>
+                          <th className="px-4 py-2">阶段</th>
+                          <th className="px-4 py-2">图片</th>
+                          <th className="px-4 py-2">标注</th>
+                          <th className="px-4 py-2">类别分布</th>
+                          <th className="px-4 py-2">入库时间</th>
+                        </tr>
+                      </thead>
+                      <tbody>
+                        {pack.batches.map((b: AdasBatchEntry) => (
+                          <tr key={b.batch} className="border-b border-gray-100 hover:bg-gray-50">
+                            <td className="px-4 py-2 font-mono text-xs">{b.batch}</td>
+                            <td className="px-4 py-2">{b.task}</td>
+                            <td className="px-4 py-2">
+                              <Badge variant={b.stage === "ingested" ? "success" : "default"} size="small">{b.stage}</Badge>
+                            </td>
+                            <td className="px-4 py-2">{b.images}</td>
+                            <td className="px-4 py-2">{b.labels}</td>
+                            <td className="px-4 py-2">
+                              <div className="flex flex-wrap gap-1">
+                                {Object.entries(b.class_counts).map(([cls, n]) => (
+                                  <span key={cls} className="inline-flex items-center gap-1 px-1.5 py-0.5 rounded bg-blue-50 text-blue-700 text-xs">
+                                    {cls} <span className="font-medium">{n}</span>
+                                  </span>
+                                ))}
+                              </div>
+                            </td>
+                            <td className="px-4 py-2 text-xs text-gray-400">
+                              {b.ingested_at ? new Date(b.ingested_at).toLocaleString("zh-CN") : "—"}
+                            </td>
+                          </tr>
+                        ))}
+                      </tbody>
+                    </table>
+                  ) : (
+                    <div className="px-4 py-8 text-center text-gray-400 text-sm">暂无批次</div>
+                  )}
+                </div>
+              ))}
+            </div>
+          )}
+
+          {/* Empty state for forward */}
+          {domain === "forward" && (!cat?.adas || Object.keys(cat.adas).length === 0) && (
+            <div className="card text-center py-12 text-gray-400">
+              <p className="text-lg mb-2">暂无 ADAS 前向数据</p>
+              <p className="text-sm">请先在送标工作台中标注并入库，或在数据湖放入 quaternion_json 格式的标注数据</p>
+            </div>
+          )}
+
+          {/* DMS/Forward: Charts — only for DMS domain */}
+          {domain === "dms" && taskEntry && isDmsCatalogScope(scope) && (
             <div className="rounded-xl border border-gray-200 bg-white p-3 mb-4">
               <div className="flex flex-wrap gap-2 items-center mb-2">
                 <span className="text-base font-semibold">{activePackTask?.label || taskEntry.label || scope.task}</span>
@@ -253,8 +321,8 @@ export const CatalogPage: React.FC = () => {
             </div>
           )}
 
-          {/* Empty state */}
+          {/* Empty state for DMS */}
-          {domain !== "lane" && !taskEntry && (
+          {domain === "dms" && !taskEntry && (
             <div className="card text-center py-12 text-gray-400">
               <p className="text-lg mb-2">暂无 {DOMAIN_TABS.find((t) => t.key === domain)?.label} 数据</p>
               <p className="text-sm">请先在送标工作台中扫描入库，或检查数据集链接是否正确</p>
@@ -266,8 +334,8 @@ export const CatalogPage: React.FC = () => {
             </div>
           )}
 
-          {/* Pack table */}
+          {/* DMS pack table — only for DMS domain */}
-          {domain !== "lane" && isDmsCatalogScope(scope) && taskEntry && tablePacks.length > 0 && (
+          {domain === "dms" && isDmsCatalogScope(scope) && taskEntry && tablePacks.length > 0 && (
             <div className="rounded-xl border border-gray-200 bg-white overflow-hidden">
               <table className="w-full text-sm">
                 <thead><tr className="bg-gray-50 text-left border-b border-gray-200"><th className="px-4 py-2">训练包</th><th>启用</th><th>train</th><th>val</th><th>test</th></tr></thead>

platform/web/src/modules/labeling/pages/ExportPage.tsx

@@ -19,6 +19,7 @@ export const ExportPage: React.FC = () => {
   const [error, setError] = useState<string | null>(null);
   const [info, setInfo] = useState<string | null>(null);
   const [importingId, setImportingId] = useState<string | null>(null);
+  const [exportingId, setExportingId] = useState<string | null>(null);
   const [buildingId, setBuildingId] = useState<string | null>(null);
   const [search, setSearch] = useState("");
   const [stageFilter, setStageFilter] = useState("");
@@ -60,8 +61,11 @@ export const ExportPage: React.FC = () => {
   const reloadCurrent = useCallback(() => load(offset, limit), [load, offset, limit]);
 
   const handleExport = async (campaignId: string) => {
+    setExportingId(campaignId);
+    setError(null);
     try { await hsapApi.labelingExport(campaignId); setInfo("导出任务已提交"); reloadCurrent(); }
     catch (e) { setError(String(e)); }
+    setExportingId(null);
   };
 
   const handleSubmitBuild = async (b: LabelingBatchRow) => {
@@ -135,6 +139,7 @@ export const ExportPage: React.FC = () => {
             <ExportBatchTable
               batches={batches}
               importingId={importingId}
+              exportingId={exportingId}
               buildingId={buildingId}
               onExport={handleExport}
               onImportVendor={handleImportVendor}

workflow.registry.yaml

@@ -55,9 +55,9 @@ projects:
     # 数据湖：宿主机 DATA/data/送标/adas → inbox/{task}/{batch}/images/
     registry: adas.registry.yaml
     packs_registry: data_packs.yaml
-    base_pack: adas_moon3d_v1
+    base_pack: adas_v1
     active_packs:
-      - adas_moon3d_v1
+      - adas_v1
     merge:
       train: lists/train_stems.txt
       val: lists/val_stems.txt